A cognitive model for understanding graphical perception.

Abstract

New advances in animation, scientific visualization, and graphical user interfaces make it essential that graphic design have advice about how to provide interfaces that mesh with users capabilities. The purpose of the dissertation is to gather what we know from cognitive psychology about graphical perception and put it in a computer program useful for designers of graphical displays. A computer program, UCIE (an acronym for Understanding Cognitive Information Engineering), simulates the mechanisms that underlie graphical perception. UCIE predicts the time it will take someone to answer a question from a graph or table. UCIE assumes the viewer follows a logical path of eye movements, called a semantic trace, to retrieve information from the graphical display. UCIE predicts total time by adding the times from the individual perceptual and cognitive components. Graph type and question type determine which component tasks are involved. UCIE moves beyond the current level of understanding by providing more detail about potential interactions among the component tasks in visual perception and cognition. UCIE has undergone rigorous empirical verification. Reaction times to yes/no questions have been gathered from 28 subjects. Each subject participated in eight replications, viewing three kinds of graphs (bar, line and tables), each with and without color and grid lines, answering three types of questions (point reading, comparisons, and trends) with two levels of difficulty. The largest predictive component is the number of fixations, explaining 31 percent of the variation by itself. Overall, UCIE explains about 37 of the variation in reaction times. Graphics reduce cognitive overhead and enable people to perform certain tasks that would be more difficult, if not impossible, when the knowledge contained in the graphic was structured in a different format. The second study assesses decision making performance using a simulated business task with three levels of task complexity and two levels of quality of a graphic decision aid. The results suggest that the quality of the graphic decision aid influences performance only when the task is very complex. The explanatory power of UCIE can be exploited in several practical applications. The graphical perception and cognition "core" of UCIE can be a prototype for the analysis of more complex displays such as those for computer assisted software engineering tools, nuclear control rooms and cockpit design.